- Talk track
- ForgeLayer is the control plane for AI software teams. It does not replace GitHub, Codex, Claude Code, Cursor, or human reviewers. It controls and records AI-generated software work before it reaches production.
- Demo action
- Open the dashboard and point to the Diff -> Policy -> Decision flow.
- Viewer takeaway
- ForgeLayer is a control layer around AI-generated software work, not another coding agent.
- Safety note
- ForgeLayer analyzes, records, and recommends; it does not execute agents.
Open demo step- Talk track
- Private beta starts in safe mode: no live agents, no autonomous workflows, no billing, no production workspace controls, and GitHub posting disabled by default.
- Demo action
- Open onboarding and show the safe pilot path.
- Viewer takeaway
- The reviewer understands what is intentionally disabled.
- Safety note
- GITHUB_COMMENT_POSTING_ENABLED should remain false unless a controlled test is approved.
Open demo step- Talk track
- The GitHub App setup is selected-repo and private-beta only. It verifies webhooks, fetches PR diffs, reads repo-owned forgelayer.yml policy, and creates safe review-run summaries.
- Demo action
- Open GitHub App setup and walk through the checklist.
- Viewer takeaway
- A team can pilot ForgeLayer on one approved repository without public self-serve rollout.
- Safety note
- Posting remains disabled by default and repos should stay allowlisted.
Open demo step- Talk track
- Review Runs are the first AI Work Ledger records: safe summaries of PR review work, risk, policy context, lifecycle status, GitHub Check Preview, and PR Comment Preview v2.
- Demo action
- Open Review Runs, filter if useful, and inspect one safe detail page.
- Viewer takeaway
- ForgeLayer creates a system-of-record summary without exposing raw diffs or raw repo rules.
- Safety note
- Review Runs omit raw diffs, raw repo rules, webhook payloads, tokens, keys, prompts, and secrets.
Open demo step- Talk track
- PromptForge v1 reviews prompts before agents create code. It scores quality and security, suggests a safer rewrite, and generates short security tests.
- Demo action
- Open PromptForge, load or paste a prompt, and show scoring, rewrite, and security tests.
- Viewer takeaway
- ForgeLayer can improve prompts before risky agent work starts.
- Safety note
- PromptForge is deterministic and does not persist raw prompt text.
Open demo step- Talk track
- AgentOS Foundation is planning/control only. It compares agent profiles, recommends routing, drafts approval checklists, and explains oversight needs.
- Demo action
- Open AgentOS and show Task Router, Task-to-Agent Workflow, and Agent Comparison.
- Viewer takeaway
- ForgeLayer can plan who should handle work before any agent is trusted with it.
- Safety note
- AgentOS does not call external agent APIs, execute agents, or run autonomous workflows.
Open demo step- Talk track
- Agent Work Records document what an AI agent was asked to do, which files may be affected, what evidence exists, which risks remain, and whether approval is still missing.
- Demo action
- In AgentOS, create or inspect a draft Agent Work Record and its ledger preview.
- Viewer takeaway
- AI work can be documented before merge readiness is trusted.
- Safety note
- Draft Agent Work Records are not production-persisted yet.
Open demo step- Talk track
- The AI Work Ledger foundation connects review runs and draft agent work records into a safe audit trail for AI-assisted software work.
- Demo action
- Open Ledger and show review-run and derived agent-work ledger entries.
- Viewer takeaway
- ForgeLayer is moving toward a system of record for AI software work.
- Safety note
- Full production ledger persistence and analytics are not live yet.
Open demo step- Talk track
- Private beta feedback is preview-only. It structures reviewer feedback and pilot interest for manual follow-up without saving data or sending webhooks.
- Demo action
- Open Feedback and create a local preview summary.
- Viewer takeaway
- The reviewer has a clear way to tell you what was valuable or missing.
- Safety note
- Feedback preview does not persist, email, or call external webhooks.
Open demo step- Talk track
- Close by asking whether the reviewer would trust ForgeLayer on a real repo if GitHub integration, workspace auth, and persistence were production-ready.
- Demo action
- Return to onboarding or beta guide and agree on the next safe pilot action.
- Viewer takeaway
- The next pilot step is explicit and does not overclaim production readiness.
- Safety note
- Do not claim billing, live agents, autonomous workflows, or production workspaces are live.
Open demo step